Skip to Content
Hägvall Labs

Mask Personal Data Before It Reaches Your AI.

Maskera finds and masks names, personal identity numbers, addresses and phone numbers in text, before it is used in AI systems, logs or analytics tools. Built for Swedish text and Swedish personal data.

Visit maskera.devGet in Touch

Your textExample in Swedish

Anna Lindqvist (900101-0000) på Verkstadsgatan 12 i Malmö undrar om sin faktura.

What the AI model sees

[NAMN_1] ([PERSONNUMMER_1]) på [ADRESS_1] i [PLATS_1] undrar om sin faktura.

Where Maskera Fits

Your text

Anna Lindqvist, 900101-0000

Maskera

  • AI Systems

    Send prompts and documents to AI models without leaking names, personal identity numbers or contact details.

  • Logs

    Keep your logs useful for debugging, minus the personal data.

  • Analytics

    Feed analytics and BI tools with text that is free of personal data.

What Maskera Detects

Each item is swapped for a placeholder like [NAMN_1] or [PERSONNUMMER_1]. The key to restore them stays on your side, so the AI model’s answer can be filled back in with the real values, locally.

Free Text

My own 43 MB Swedish AI model catches what rules never can: names, places, organizations and street addresses, even in lowercase, all caps and genitive forms, the way people actually type.

  • Names
  • Places
  • Organizations
  • Street addresses

Structured Data

Rules with format checks and checksums, so something like 2019-2024 is never mistaken for a bankgiro number. Your own formats, like case or customer numbers, join as rules of their own.

  • Personal identity numbers
  • Coordination numbers
  • Organization numbers
  • Phone numbers
  • Email
  • Postal codes
  • Apartment numbers
  • Bankgiro
  • Plusgiro
  • IBAN
  • Card numbers
  • IP addresses
  • Links

How It’s Built

  • Self-Hosted by Default

    Maskera is installed on your side and the text stays on your side. No cloud dependencies, no third party in the flow.

  • Built for GDPR

    Masking personal data early is data minimization for real: less data to protect, fewer systems in scope.

  • Simple to Integrate

    A straightforward API that fits the flows you already have: before the AI call, before the log write, before the export.

Inspect It Yourself

The core of Maskera is open: the code is on GitHub, the package on npm and the NER model on Hugging Face. You don’t have to take my word for how it works, you can read it and run it before you talk to me.

See Maskera in Practice.

How accurate is it? That’s for the pilot to answer, not a sales deck. I run pilot projects with companies and organizations in Sweden. Get in touch and I’ll show you what it looks like, or read more at maskera.dev.

Get in Touch